Hi all, The notebook migration tool tracked under 4301 <https://github.com/apache/texera/issues/4301> is now on main! It converts a Jupyter notebook into a Texera workflow using an LLM, and keeps a mapping between notebook cells and the operators generated from them, so you can see which cell produced which operator while you edit. Big thanks to Meng Wang for reviewing the project and Professor Li for his advising!.
It is off by default behind the pythonNotebookMigrationEnabled flag. >From a user's point of view: - Upload a notebook from the workspace toolbar and pick a model from the ones the deployment exposes. The browser calls Texera's LiteLLM proxy with the configured deployment's API key to do the conversion. - The workspace reloads with the generated workflow on the canvas and the notebook open read-only in a panel beside it. - Generated operators are PythonUDFV2, so the result is an ordinary workflow that can be edited, run, versioned and shared like any other. - The notebook and the mapping are persisted per workflow, so reopening a workflow reopens its notebook. I wrote up the design and the challenges along the way on our blog page, go here if you want to learn more!: https://texera.apache.org/blog/2026/09/15/notebook-migration-tool/ Feedback is welcome, particularly from anyone who runs it against a real notebook and can share their experience with the tool. Support for Python scripts and R is in progress. Thanks, Ryan
